Human vision guides lane keeping and hazard anticipation during driving. However, isolating how visual field constraints affect steering is difficult in real driving. This study used immersive VR with a depth-aligned aperture to restrict vision while participants drove curved roads under several conditions. Results revealed that misaligned restrictions impaired steering, while tangent point alignment partly improved performance. Results highlight how VR can probe visual–motor mechanisms in driving.

Despite the success of Siamese networks in image indexing, face recognition, and signature verification, there has been little research on designing their architectural space compared to convolutional neural networks (CNNs). This work aims to automate the design process of Siamese network architectures and improve their performance in tasks that involve image similarity computing such as indexing and retrieval. To achieve this goal, in contrast with the current literature that focuses on improving the design of the backbone CNN, we use Differentiable Neural Architecture Search (DNAS) to explore the architecture of the Multi-Layer Perceptron (MLP) component of siamese networks, namely the projector and/or predictor heads. The main objective of these MLPs is to enhance the ability of backbone CNNs to learn strong representations from unlabeled data. Using a well-known contrastive learning framework (SimCLR) as a baseline, we show that our approach managed to improve performance on several computer vision tasks such as image classification (ImageNet) and content-based image retrieval (INRIA Holidays).

Action segmentation is a challenging task that requires accurate parsing and labeling of each action. There are two types of methods for action segmentation. The first type primarily focuses on extracting high-quality features from videos, while the second type focuses on combining textual and perceptual features through multimodal fusion. However, both types of methods have their limitations. The first type is limited to a single modality and does not leverage multimodal information, while the second type, although promising, is restricted by the language used to describe the actions in the texts. To solve these problems, we propose in this paper, an Action Text Diffusion Prior Network (ATDPN) which simultaneously improves the quality of the extracted visual features (by introducing a Video-level Diffusion Prior Sampling) and integrates the textual information to fullest extent. This leads to superior action segmentation results. Our experiments performed on GTEA dataset demonstrate the effective feature extraction ability of ATDPN.

This paper proposes a simple, model-independent method to estimate the sideslip angle of Powered Two-Wheeled Vehicles (P2WV) using knowledge about the road model and an Inertial Measurement Unit (IMU). This proposed model is then tested using simulator software (BikeSim) on different scenarios with different tracks, velocities, and sample rates. Our results indicate the validation of the model. With these promising results, we propose some future applications where this work can be utilized.

In this paper, the authors propose a visual-inertial algorithm to estimate the lateral velocity of a motorcycle traveling at high speed along a single-carriageway road. The approach comprises the following steps. First, a monocular camera captures real-time images of the road ahead. Lane markers present in the image are detected and segmented using image processing techniques. Next, a bird's eye view transform is applied, and the dashed center lane markers are isolated. The motion of these markers is computed using an image registration algorithm and is expressed in the motorcycle body frame using orientation estimates from an inertial measurement unit. Finally, this measurement is combined with readings from an accelerometer using a Kalman filter to produce a filtered estimate. The approach was validated using data from simulations of two scenarios created in the BikeSim simulation software suite. In the first scenario, the motorcycle performs a double lane-change across both lanes of a straight road. In the second, the motorcycle navigates an s-shaped bend.

This paper deals with an observer-based identification framework to estimate both lateral dynamics states and tires’ cornering parameters in the perspective of designing advanced rider assistance systems for powered two-wheeled vehicle. An adaptive observer is proposed to reconstruct the state variables regardless of the forward velocity variations and to estimate the real unknown tires’ parameters. The stability and convergence analysis of the proposed observer is based on the Lyapunov theory, the persistency of excitation and the general Lipschitz condition. To enable this observer design, the linear parameter varying observer is transformed into Takagi-Sugeno exact form where the sufficient conditions are given in terms of linear matrix inequalities. Finally, an evaluation framework is proposed to provide a critical overview of the method’s effectiveness. The proposed adaptive law is compared to direct estimation and dynamic inversion estimation methods. Co-simulation scenarios are performed by using both BikeSim© motorcycle simulator and real data-log obtained from an instrumented electrical scooter.

The purpose of the present work is the reconstruction of motorcycle lateral dynamics. The main idea is to estimate pertinent states and unknown inputs (rider action) with respect to nonlinear outputs due to motion transformation frames (inertial sensors are away from the local frame). To overcome this issue, we propose a new Unknown Input Observers with variable output matrix. In this paper, we take into account the ground truth measurements provided in the body-fixed frame, parametric uncertainties as well as sensors noise. This step leads to a nonlinear parameter-dependent output equation with unmeasured premise variables in the observer design. The observer synthesis is specified in term of convergence and stability study by considering a quadratic Lyapunov function associated with the Input To State Stability (ISS) property. Sufficient conditions are agreed in terms of Linear Matrix Inequalities (LMIs). Finally, the performances, usefulness and robustness of the proposed approach are assessed throughout an electric Scooter under urban riding scenario.

This paper presents an image-based approach to estimate the motorcycle roll angle. The algorithm estimates directly the absolute roll to the road plane by means of a basic monocular camera. This means that the estimated roll angle is not affected by the road bank which is often a problem for vehicle observation and control purposes. For each captured image, the algorithm uses a numeric roll loop based on some simple knowledge of the road geometry. For each iteration, a bird-eye-view of the road is generated with the inverse perspective mapping technique. Then, a road marker filter associated with the well-known clothoid model are used respectively to track the road separation lanes and approximate them with mathematical functions. Finally, the algorithm computes two distinct areas between the two-road separation lanes. Its performances are tested by means of the motorcycle simulator BikeSim. This approach is very promising since it does not require any vehicle or tire model and is free of restrictive assumptions on the dynamics.

This paper presents a vision-based approach to prevent dangerous steering situations when riding a motorcycle in turns. The proposed algorithm is capable of detecting under, neutral or over-steering behavior using only a conventional camera and an inertial measurement unit. The inverse perspective mapping technique is used to reconstruct a bird-eye-view of the road image. Then, filters are applied to keep only the road markers which are, afterwards, approximated with the well-known clothoid model. This allows the prediction of the road geometry such as the curvature ahead of the motorcycle. Finally, from the predicted road curvature, the measurements of the Euler angles and the vehicle speed, the proposed algorithm is able to characterize the steering behavior. To that end, we propose to estimate the steering ratio and we introduce new pertinent indicators such as the vehicle relative position dynamics to the road. The method is validated using the advanced simulator BikeSim during a steady turn.

Recent advanced riding assistance safety systems, such as electronic yaw stability control, adaptive cruise control, and lane-keeping systems, require good approximations of motorcycle inertial properties, such as yaw, roll and pitch moment of inertia, this parameter can vary significantly with the rider's weight and heavy baggage placed on the luggage rack. This paper presents further research on parametric identification of two wheeler vehicles, carried out using a recursive Levenberg-Marquardt parameter identification formulation. This approach needs the use of sensitivity functions to identify acceleration responses in time domain by updating coupled inertial parameters value. The identification algorithm is implemented in MATLAB / Simulink software. Data and prior value are taken from the professional motorcycle simulation software BikeSim (based on high fidelity virtual motorcycle models).

The flatness-based approach is presented in this paper in order to estimate the motorcycle lateral dynamics such as the roll angle, the lateral tire forces or the steering torque from basic measurements. The model of the motorcycle associated with the flatness theory are used to express the unknown states and input in terms of nonlinear functions depending only on the measures and their time derivatives up to a given finite order. These time derivatives are estimated via a non-asymptotic differentiator. Finally, the ability of the proposed observer to estimate the motorcycle dynamics is illustrated through two simulation scenarios performed with the well-known motorcycle simulator "BikeSim".

This paper highlights the necessity of the rider weight consideration during observer's design for motorcycle dynamics estimation or control. It presents a novel approach using a linear parameter varying (LPV) model associated with the well-know Takagi-Sugeno (TS) methods to derive a robust observer regarding the rider weight uncertainty. Then the proposed solution is illustrated with an application to a proposed observer in our previous works by comparing results of estimation between a nominal, a heavier and a lighter rider. Finally, a complete simulation scenario shows the ability of the proposed method to estimate the lateral motorcycle dynamic states considering an uncertain rider weight.

Powered Two-Wheeled vehicles (PTWv) are an increasingly popular means of transport. The cost and the risks of the development phase of this vehicles has to be diminished in order to ensure acceptable levels of comfort and safety for riders upstream of hazardous driving situations. It is required to study motorcycle while riding in cornering and lane-keeping to interpret the dynamics behavior. Thus, we need to obtain dynamic model that tightly adjusts to the real lateral behavior of the motorcycle, in the way that it will lead to precise simulation and experimental results. A technique for cascade identification of parameters based on data and optimization algorithm, is presented here. This methodology makes profit of the possibility given by this type of algorithm for solving multiple objectives function consecutively. After the identification method is outlined, simulations and experimental results are presented in order to confirm the accuracy of the parameters estimation under the persistent condition of the inputs.

This paper deals with modeling, control and identification issues of 2 degrees of freedom (DOF) low cost driving simulator validated by experimental results with a human in the loop. A first study, dealing with choices of the platform's motion, has conducted us to an original architecture allowing the restitution of longitudinal and yaw movements based mechanics. To better immerse the driver in the virtual world and stimulate his perception, a haptic feedback steering wheel is be implemented to assist the human in driving. The present motion platform is designed to help psychophysicists to assess the effects of yaw components on the simulator sickness. Experimental studies were carried out to devise a characterization of the platform capabilities, frequency responses and for classical drive operation. Conclusion and future works are given.

This paper deals with motion control problem for a 2 DOF small driving simulator. The main idea is to test and compare performances of different washout algorithms applied to such platform category. The experimentations allow us to have the best compromise between quality of the perception (sensation), implementation complexity and platform architecture. Implementation of different washout algorithms (optimal, adaptive and classical one) are discussed. Only the longitudinal restitution is studied. The results show that there is not significant differences between these approaches using with platform type. The lack of pitch DOF in our simulator does not allow a restitution of the sustained acceleration and no coordination between longitudinal and pitch channels may be done